基于用户多属性与兴趣的协同过滤算法*

传统的协同过滤算法广泛应用于推荐系统领域,但该算法仍存在用户冷启动和数据稀疏性问题,造成算法的推荐质量较差。对此,提出一种基于用户多属性与兴趣的协同过滤算法AICF(Attributes and Interests Collaborative Filtering)。首先通过对多种用户属性分配权重计算出用户多属性相似度。其次利用改进的Slope One算法填充用户-项目评分矩阵,然后计算基于隐性标签的用户兴趣相似度。最后基于两种相似度的组合进行推荐。实验结果表明,AICF算法不仅明显提高了推荐结果的准确性,同时也解决了用户冷启动和数据稀疏性问题。

Collaborative filtering algorithm based on multiple attributes and interests of users

The traditional collaborative filtering algorithms have been widely used in the field of recommender systems. However, there is a decline in the quality of recommendation due to user cold start and data sparsity. Therefore, this paper proposes a collaborative filtering algorithm called AICF (Attributes and Interests Collaborative Filtering), which is based on multiple attributes and interests of user. Firstly, AICF assigned a variety of user attributes weights to calculate the user multi-attribute similarity. Secondly, AICF applied the improved Slope One algorithm fill user-item rating matrix, and then implicit tag similarity of user interest can be worked out. Finally, AICF combined these two similarities to get recommendation results. Experimental results show that the proposed algorithm not only highly improves the accuracy of the recommendation, but also improves problems of user cold start and data sparsity.